Evaporation Mechanisms for Particle Swarm Optimization

摘要

This paper presents a novel approach to dealing with sample noise in Particle Swarm Optimization (PSO) by introducing a heterogeneous swarm whose particles have different evaporation factors. So far, previous works have considered only homogeneous swarms in which the evaporation factor is the same across particles. However, choosing a proper factor largely depends on the severity of noise in the optimization problem. If the level of noise cannot be determined a priori, arbitrarily choosing the evaporation factor can lead to rather poor results. This paper shows that heterogeneous swarms are generally better than homogeneous ones in low to medium levels of noise, and also in its absence.